Final Thoughts

I am super happy with the outcome. It is definitely the best voice work I have done, I didn't really know I was capable of that until I put everything into it. As a film trailer style short I think it really sells the scene I was trying to set, and paints the picture just how I wanted it, leaving the audience hopefully keen to see the movie itself. What will Manny do?

The processing app is quite general purpose, I had a ball making it. Probably the most fun thing to do is grab image URLs from the web and bung them right into the Open File dialog box that pops up when you run the app, and seeing what you can do to these images. Let me know if you are interested in seeing it (or making some hex graphics with it) since I didn't hand it in. It's a fun tool for exploring a similar geometric minimalism to pixel art, yet it reads so differently and its elementary 'language' is so different.

Feedback

Showing Elle the final video, she commented about the flatness of the effect overlay and how it could be more convincing if the skin textured showed through. Since I wanted to do this anyway from the beginning, I decided to revisit it. I actually found that the sequence I was importing from Premiere had less than 100% opacity, so in After Effects the max opacity I was working with wasn't actually black as it should have been, leading to a much too light result when using some blend modes. I fixed this up and used the overlay blending mode, and the effect definitely saw an improvement in believability.

Audio

I had wanted to use audio fairly early on to give the piece more mood and backstory, since there isn't time to convey the things I need through any other means. I also didn't want to over-explain, since making the audience intrigued to see more is a major goal of mine.

The format that made the most sense was disembodied voices, echoing in Manny's mind as he checks how much time he has left in the mirror for the millionth time. Manny would be distraught and his mind would be a mess of panicked thought, memories and emotions, so I wanted to convey that inescapable feeling, the feeling that these Mafia men were hounding him even now and he can't run away no matter where he goes.

I considered finding some samples, but quickly gave up on that idea in favour of recording my own, not wanting to have copyright issues, and also for my own satisfaction of having as much of this work as possible be my own. Had time not been a factor I would have done my own music too, since I love that, but it gets impractical to go that far. So I shut myself in my room with a Skype headset, some reasonable headphones and a copy of Audacity, and set to work.

Finding the characters was the first challenge. I wanted a mixture of different people, sounding like snippets of scenes we haven't seen yet, memories of conversations (or 1-sided conversations where perhaps Manny was in no position to speak). I wanted to focus on the bad guys, but ended up slipping a sympathetic female voice in the mix to sort of say what Manny would be thinking. I had a few lines thought up, but the lines seemed to come organically through trying take after take to get a convincing criminal character going. I must have done hundreds of takes just talking in a particular voice, trying to iron out vowels or tones that didn't fit with the next try. I learned a valuable trick just by experimenting, it was crucial to making that mobster-sounding guy to actually distort my mouth a lot while speaking.

The effects I used to push the sound further from just sounding like me were what closed the deal. Accents and mouth-mangling can get you so far, but you are physically working with a different set of pipes than some people. So the pitch changer was employed carefully to make me bigger and scarier, or in one case female. The noise removal was essential to getting a clean recording, and was a huge help. The reverb topped it off, and required a lot of fiddling (reverb has many parameters!) but in the end I feel like I nailed the 'surfacing memory' sound without washing it out too much and adversely affecting clarity. I also used the DeEsser in Premiere, although its effect was quite subtle.

For the backing music, I knew I needed something sombre and building, like a serious movie trailer, so I trawling Free Music Archive for a while. Eventually, after deciding against choral music (conflicts with spoken audio which is more important) I found a lovely bit of soundtrack work by Kevin MacLeod called 'Devastation and Revenge'. Perfect. Even more perfect was the length of its moody intro, before suddenly arresting the listeners attention right when I cut to black. With minimal fiddling it slotted right in! This piece really caps off the mood of the video for me, colouring how the audience interprets the voices and providing a great crescendo.

Compositing

The process I tend to use is:

1. Cut the base footage in Premiere
2. Compile the frames from Processing into a Sequence in Premiere
3. Import the two into After Effects to composite and do video effects
4. Import the AE comp into Premiere again to do audio arrangement and effects, and fade ins/outs
5. Render the complete Premiere sequence with Media Encoder

Tracking in AE was harder than expected, due to slightly out of focus bits of footage I hadn't spotted, and a lack of tracking marks. I didn't want to use tracking marks, because artificially smoothing my skin would be noticeable seeing as my skin is far from smooth normally. I also wanted the imperfections in my skin to show through the overlay to make it more convincing. I managed okay with a few different attempts though, using things like the corners of my nose or eye.

I tried out all the blend modes for the overlay layer, but there was an issue with the effect getting too light and being unable to read it, or it not standing out as I wanted it to, so I reverted back to the layer blend mode and used some masking with wide feathering to replicate the main lighting on my face.

Once that was done, I fiddled around with different video effects just to experiment, and found that it looked much moodier in black and white. This suit the noir/prison sort of feel I wanted so well that I switched it to black and white by using a bunch of color correction effects (the contrast is never good enough by default when switching to B&W).

Filming

During filming, I decided to change the position of the overlaid effect from my neck to my forehead. This was due to scene lighting conditions making the forehead much easier to focus on.

Instead of touching my finger to my neck, I now touch it to my temple to activate the visualisation.

Enter Processing

Yes! I found a way to include a Processing application as part of this project, much like my esteemed peer Ben Richards. While fiddling with Illustrator trying to figure out a way to compose hexagon-based forms quickly, it hit me than converting regular symbols and imagery into hexagons is something Processing could probably do for me, and with far more flexibility. Even with the potential for animation! So I set about writing this very app.

I call it Hexagonize. There is not much agonizing involved, so its not a clever pun or anything, but not everything can be a clever pun sadly.

The essence of the app is to take an input image, sample hexagonal areas of that image (that tessellate), and then place hexagons there. I actually cheat, and just use squares that approximate the hexagons, but it makes little difference. Early versions had a brightness threshold, and if the average brightness of all pixels in that area was low enough it would plonk a black hex down there. This made for interesting symbols, but it was poor for images, so I improved it first to draw a greyscale hex representing the average brightness.

Don't judge my choice of images, if you have to stare at pixels a lot when writing image analysis apps, it's nice to look at something pretty. :D

I moved on from this to colour, and accidentally learned something about greyscale in the process. It turns out that HSB colour systems are terrible at perceived brightness, so just modifying the saturation value of a colour over a whole image will give you a much worse (less contrasty) result than the Hue/Saturation/Lightness adjustment tools that Photoshop-like applications give you. They are doing different maths to juggle the other values when you shunt that Saturation slider all the way to greysville.

Anyway.

I added controls for the hex edges as well, because for some images you want to really know its hexs, and for others they look great borderless. There are controls for moving the hex field about to, just so you can force it to resample in a slightly offset location (I thought this might come in handy later).

Yeah I had fun plugging random images from the internet into it.

When I started trying to apply this hexagonising to the actual images I wanted to use as my effect, I ran into some issues with text readability. Everything looks cooler in low res, but text needs to be mighty large.

I chose then to simplify further, and have the graphic just be the number 16, but have that number be very interesting to look at, and arrive in an animated way. I had built animation potential into the program already, so that wasn't too bad. Basically I just set the hex field to automatically drift left, and it re-samples every frame. The shimmery effect from this is very interesting on some images at some resolutions, worth exploring another time I think. The most interesting accident from playing with this is illustrated in the animated GIF below.

With any image I had containing rain, it seemed to animate the rain. This works even in images containing people and other objects which shouldn't be moving, the rain animates around them, while their edges only shimmer. I suspect this is just to do with how our eyes read images, and the size of colour blocks, details, edge detection, feature size, that kinda thing. The best example I ran into of this effect containing a person is a little *ahem* risque, so I won't post it here, but ask me if you are curious.

Hexagons Arrive

Doing a very quick test of putting a vector graphics overlay on top of a random face image, I noticed something immediately. The text has edges that are just too sharp. This can be fixed by blending/blurring, but there is just something so wrong looking about it being hi-res. If this is being created by a virus, shouldn't it look more organic?

Thinking further on this I decided to break the basic graphics up into modular bits, so it would look more like the virus is cobbling these images together from bits. To maintain the high-tech angle, I thought hexagons would go nicely. Hexagons and me have gotten along well lately.

Decision: Countdown

I decided to go with the Countdown concept, as people seemed to react well to its deeper story, identifying with its character. It also requires less props and tricky shots, allowing for more polish and a focus on mood, acting and audio.